Re: Constraint failure in rattle algorithm for some atoms, simulation becomes unstable

From: Brian Radak (brian.radak_at_gmail.com)
Date: Fri Mar 22 2019 - 11:20:26 CDT

1) It doesn't look like you actually did any minimization.
2) I very often find it useful to extend my PBC boundary since poor
wrapping and ragged edges of newly solvated structures can cause bad steric
clashes at the boundary. Usually adding 0.5-1 A in each direction works.
This is due to most solvator plugins (tleap and VMD both) yielding
notoriously under dense water boxes.

Some notes for the community and not necessarily directed towards your
specific problem:

1) There is generally no reason to set xstFreq - unless you plan on using
high frequency box output I strongly recommend not using this *at all. *This
appears to be a bad practice carried over from CHARMM-GUI inputs.

2) Setting high frequency restarts is also a very bad idea and can
*severely* impede performance. restartFreq should really never be below
tens of thousands of steps (the order of a restart every few hours of human
time).

3) IT IS NOT RECOMMENDED TO ADJUST stepsPerCycle! This probably does not do
what you think it does and its meaning has been incorrectly reported in
multiple places (it's not the same thing as the pairlist frequency). The
default of 20 is almost assuredly the best choice.

4) I don't understand the community's fixation with rapidly heating up
structures (unless you are annealing). I would argue that this is an
obsolete protocol due to the routinely long equilibration times that are
now possible. An exception might be made for crystal structures obtained at
very low temperatures, but this is unusual.

Some other relevant notes to 4:
   -- Make sure you use "reinitvels" after minimization - it is an
unfortunate quirk of NAMD minimization that velocities are discarded.
   -- Setting LangevinTemp near room temperature along with any reasonable
damping coefficient will almost surely heat your system faster than an off
the shelf protocol (even if you start at 0K). You can watch the kinetic
energy time series actually get scaled DOWN by reassigning every ps or so,
even when you intend for it to be scaling UP.
   -- Slow heating a system with restraints also seems pointlessly
delicate. Strong restraints (10 kcal/mol-A^2 per atom!) are going to be
very resistant to normal temperature fluctuations.

BKR

On Fri, Mar 22, 2019 at 6:03 AM Akshay Prabhakant <
akshayresearch16_at_gmail.com> wrote:

> ## OUTPUT/INPUT
> # Amber/(t,s,x)leap generated parm and crd file
> parmfile apo_cap.prmtop
>
>
> # Input
> set temperature 300
> ambercoor apo_cap.rst7
>
> ## bin coord ##
> bincoordinates min_02.restart.coor
> #binvelocities min_02.restart.vel
> extendedSystem min_02.restart.xsc
>
> # Output
> restartfreq 1
> dcdfreq 1
> xstFreq 1
> outputEnergies 1
> outputPressure 1
> outputname eq_01
>
> ## SIMULATION PARAMETERS
> # AMBER FF settings
> amber on
> rigidBonds all
> useSettle on
> rigidTolerance 1.0e-8
> cutoff 10.0
> pairlistdist 11.0
> switching off
> exclude scaled1-4
> readexclusions yes
> 1-4scaling 0.83333333
> scnb 2.0
> zeromomentum on
> ljcorrection on
> watermodel tip3
>
> # Integrator Parameters
> timestep 1.00
> nonbondedFreq 1
> fullElectFrequency 1
> stepspercycle 1
>
> # PME settings
> PME on
> PMETolerance 1.0e-6
> PMEInterpOrder 4
> FFTWUseWisdom no
> PMEGridSizeX 54
> PMEGridSizeY 54
> wrapAll on
> PMEGridSizeZ 54
>
> ## position restraints ##
> constraints on
> consref modified_cap.pdb
> conskfile modified_cap.pdb
> conskcol B
> constraintScaling 10.0
>
> ### NVT ###
> langevin on ;# do langevin dynamics
> langevinDamping 1 ;# damping coefficient (gamma) of 1/ps
> langevinTemp $temperature
> langevinHydrogen off ;# don't couple langevin bath to hydrogens
>
> # periodic cell
> cellBasisVector1 50.8726134 0.0 0.0
> cellBasisVector2 0.0 50.8903175 0.0
> cellBasisVector3 0.0 0.0 50.8639184
> cellOrigin 0.0 0.0 0.0
>
> binaryoutput yes
> binaryrestart yes
>
> ## EXECUTION SCRIPT
> temperature 0
> seed 1010 # Initial velocity distribution is performed to meet this
> Initial temperature (in K)
> reassignFreq 1 # Number of steps between reassignment of velocities (e)
> reassignIncr 0.01 # Increment used to adjust temperature during
> temperature reassignment
> reassignTemp $temperature # The value of temperature to be kept after
> heating is completed
> reassignHold $temperature
> numsteps 100
>
>
>
> This is the equilibration script that i use for raising temperature of my
> system from 0 to 300K. It seems to be giving the following error:
>
> REASSIGNING VELOCITIES AT STEP 0 TO 300 KELVIN.
> ERROR: Constraint failure in RATTLE algorithm for atom 4043!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 49!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 2919!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 3188!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 214!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 5045!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 6347!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 4843!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 2268!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 3314!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 3329!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 3627!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom 4925!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Exiting prematurely; see error messages above.
>
>
>
> Can somebody help me in fixing this, I have tried to minimize my structure
> for a very long time too, but I still get this error. Is anything wrong
> with my script? If so, please mention the corrections. Thanks in advance.
>

This archive was generated by hypermail 2.1.6 : Wed Dec 04 2019 - 23:20:39 CST